1. Technische Universität München Application Performance Monitoring of a scalable Java web-application in a cloud infrastructure Final Presentation August 5, 2013 Student:Michael Rose Supervisor:Prof. Dr. Florian Matthes Advisor:Alexander Schneider Dr. Thomas Büchner

2. Technische Universität München Application Performance Monitoring of a scalable Java web- application in a cloud infrastructure Agenda Application Performance Monitoring APM for Tricia Evaluation Further Work Summary August 5, 20132

3. Technische Universität München Application Performance Monitoring of a scalable Java web- application in a cloud infrastructure Application Performance Monitoring (APM) „Monitoring is the process of maintaining surveillance over the existence and magnitude of state change and data flow in a system.“ 1 How fast is my application?... If the application is not fast... –When is it slow?everytime, on a certain time of the day,... –What is slow?everything, a certain type of operation, a single request,... –Why is it slow?bug in the application, insufficient resources, wrong configuration,...  Solve the problem August 5, 20133

4. Technische Universität München Application Performance Monitoring of a scalable Java web- application in a cloud infrastructure APM - Stakeholders Customers –improve their business –application must be available and usable Cloud Operations –Is there need for action? (more resources, more servers,...) –require application-level information –standardized interface Developers –Are there any Problems? –If so: What are they? How can they be solved? August 5, 20134

5. Technische Universität München Application Performance Monitoring of a scalable Java web- application in a cloud infrastructure APM – Key Metrics Overall response time Handler duration Database operation execution time Elasticsearch operation execution time Requests per Minute Database operations per Minute Elasticsearch operations per Minute System Load Process Load Memory Consumption August 5, 20135

6. Technische Universität München Application Performance Monitoring of a scalable Java web- application in a cloud infrastructure APM – Means of Evaluation Checking for problems  Performance Snapshot created at a regular interval (e.g. 2 hours) detailed information for all monitors in the system use as indicators  maximum execution time August 5, 20136

7. Technische Universität München Application Performance Monitoring of a scalable Java web- application in a cloud infrastructure APM – Means of Evaluation Verifying a Problem  Graphical Visualization create time series see if just a singular event detect dependencies August 5, 20137

8. Technische Universität München Application Performance Monitoring of a scalable Java web- application in a cloud infrastructure APM – Means of Evaluation Tracking a Problem  Performance Trace created in case of a problem gather all monitors used in the executed operation print detailed information  Find the monitor responsible for the slowdown August 5, 20138

9. Technische Universität München Application Performance Monitoring of a scalable Java web- application in a cloud infrastructure APM for Tricia Existing Solutions / Libraries –New Relic –Kieker –Java Simon –RRDTool Existing Standards –Java Management Extensions (JMX) –Nagios –collectd, ganglia,... August 5, 20139

10. Technische Universität München Application Performance Monitoring of a scalable Java web- application in a cloud infrastructure APM for Tricia - Architecture August 5, 201310

11. Technische Universität München Application Performance Monitoring of a scalable Java web- application in a cloud infrastructure APM for Tricia - Architecture August 5, 201311 Tricia Monitoring Tool Performance Snapshots Performance Traces Graphical Visualization Detection Tracking Detection Verification

12. Technische Universität München Application Performance Monitoring of a scalable Java web- application in a cloud infrastructure APM for Tricia - Architecture August 5, 201312

13. Technische Universität München Application Performance Monitoring of a scalable Java web- application in a cloud infrastructure APM for Tricia - Architecture August 5, 201313

14. Technische Universität München Application Performance Monitoring of a scalable Java web- application in a cloud infrastructure Evaluation - Context August 5, 201314

15. Technische Universität München Application Performance Monitoring of a scalable Java web- application in a cloud infrastructure Evaluation – Problem August 5, 201315 Customers reported slow performance Specific tasks took around 10 seconds to complete  Creation of a workpace Problem was reproducable

16. Technische Universität München Application Performance Monitoring of a scalable Java web- application in a cloud infrastructure Evaluation – Checking for a problem August 5, 201316 Look at Performance Snapshots NameHitsAvgTotalMax handler608292177 92085 877 entity.persist1 5411 6602 559 448154 076 es1 8865801 094 08485 545 es.searchables_production_ multitenancy.commit 1961 161227 71384 974 db221 4871317 45760 022

17. Technische Universität München Application Performance Monitoring of a scalable Java web- application in a cloud infrastructure Evaluation – Verification August 5, 201317 Graphical visualization

18. Technische Universität München Application Performance Monitoring of a scalable Java web- application in a cloud infrastructure Evaluation – Tracking Down August 5, 201318 Perform slow request Analyze Performance Trace 9 607ms for /space/submit NameHitsTotal webserver.handlerRequests19 607 handler19 605 es108 788 es.searchables_production_multitenancy.commit18 686

19. Technische Universität München Application Performance Monitoring of a scalable Java web- application in a cloud infrastructure Evaluation – Tracking Down August 5, 201319 Commit operation performed by a single method Method is basic ES library call Implementation according to documentation Research on the internet produced no results Only environment with this problem

20. Technische Universität München Application Performance Monitoring of a scalable Java web- application in a cloud infrastructure Evaluation – Results August 5, 201320 Low-level analysis revealed slow file deletion No bug in Tricia Issue caused by target infrastructure Provider was notified, working on solution

21. Technische Universität München Application Performance Monitoring of a scalable Java web- application in a cloud infrastructure Further Work August 5, 201321 Higher degree of automatization –Extract data from log files –Aerts & Notifications Even more data via JMX –Performance Snapshots –Performance Trace

22. Technische Universität München Application Performance Monitoring of a scalable Java web- application in a cloud infrastructure Summary August 5, 201322 Identification of key metrics Examination of existing solutions & standards Successfully employed APM in Tricia Evaluation of developed solution How-To for using the solution

23. Technische Universität München Application Performance Monitoring of a scalable Java web- application in a cloud infrastructure August 5, 201323 Questions & Discussion

24. Technische Universität München Application Performance Monitoring of a scalable Java web- application in a cloud infrastructure August 5, 201324 Thank you for your attention. Michael Rose